RAID has "levels" to refer to different configurations.

RAID 0 (striping) means interleaving among multiple drives to increase performance. Lose any one drive, you lose everything - or at least interleaved "stripes" will be lost from everything. So you would never use that, except for something like a cache to hold the .o files when compiling some huge thing where the build runs overnight back in the 90s.
RAID 1 (mirroring) means multiple exact copies of one drive. Lose up to the last drive, you retain everything. Read performance is theoretically higher, write performance is worse.
RAID 5 (block level striping with distributed parity) requires at least three drives. You can lose any one drive and keep everything, but you get more storage space than you would with RAID 1, especially as you increase beyond three drives, due to the tricks of being able to recover the missing data from the error correcting code.

Those were the common levels we had to memorize for things like CompTIA A+, there are a bunch more, including recursive ones like 1+0. It's all described on wikipedia.

At least for the home user, I don't think any of the performance advantages of RAID over one drive make sense any more, because you can just use an SSD.
RAID isn't a substitute for backups, since software bugs and electrical failures could cause multiple disks to fail at the same time. It also was always weird to assume failures are independent & identically distributed, when it's the same model drive living in the same environment (voltages, heat, vibration, etc.) for the same amount of time, so you'd think failures would be correlated to expect the drives to fail all at once. Also, if the sysadmin isn't paying attention, it's common for the array to enter a degraded state from a failed drive, and for no one to do anything about it until a second drive fails.

If a motherboard has onboard RAID, I now setup a 2 disk 1+0 array.

I collected about 3 shoeboxes worth of 1TB spinning rust from work so have plenty of stock.
In theory if a drive fails rather then having to do a complete reinstall, I can just insert a replacement disk.

I say theory as motherboard raid typically relies on software and have experienced first hand it's not always to be trusted.
I also notice a slight performance drop when accessing folders on the hard drive, but its a fraction of a second and games I haven't noticed any real world difference.

I don't really consider it a backup, more a convenance should a drive fail with my hard drives that have already been used in desktop pc's at work for a number of years.

My backup strategy is more like
Actual important files are on both my "server" and google drive. (the server can be a standard PC or NAS, or anything, just needs plenty of storage space.
My retro files are on the "server" as well as on the retro PC on a second partition. As I have more then 1 retro PC this means I'll actually have 3/4 copies on my network.

Does mean if my house burns down I'll loose all my retro files, but if my house burns down I wont have any retro rigs anyway (and bigger problems in my life)

I‘d like to throw in a different idea. When I was setting up my NAS with TrueNAS I was thinking about this aspect as well and, at least for me, I was lucky enough so far to not loose any data because of a disk that stopped working. However, I managed to delete some data by accident more than once. So that was a scenario I wanted to have covered.
I ended up using two disks for each of my network shares but without a RAID. One disk is used to share on the network. This is were I read and write the data from. A second disk is in the same machine but it is not shared on the network. I then used rsync with a cron job to sync these disks once a week. Like this deleting data by accident does not mean that it is lost, at least not until the next synchronization happens, while on the other hand the worst case scenario is that I loose one week of data in case a the disk dies that is shared on the network. That felt like a good comprise to me.
I now use ZFS with snapshots to achieve the same thing but as this is a more advanced solution, maybe the option to sync in an automated way is an option to consider for you.

For the really important data I have a raspberry pi with a USB disk at my parents place where I push backups to.

I also use ZFS with snapshots to protect me from accidental deletions. It works great and it is super convenient: snapshots can simply be mounted temporarily so that you can retrieve any file that was lost.

I use RAID on my NAS servers for mirroring. I have had drives fail over the years so the redundancy has been worth it.

I don't use RAID on my desktops, since anything important I store on my NAS.

Ive used RAID disks since the early 2000s. Ive still got all the data I had back then today because the RAID I had protected it. Had it been on a single disk I would have lost the lot a long time ago because single drives fail.

I run mdraid with a raid6 array with all my live data and archives and stuff on.

Was a raid5 before, but once had a close call with disk failures, so I upped it to a 6.

Kernel mdraid is very robust and versatile. Can do size and layout changes on the fly while the volume is mounted read/write.

Weekly scheduled scrubs make sure all the sectors on all the drives read properly and checksums check out.

This is no substitute for a backup, but surely a good way to keep the stuff you want available, available on demand.

A backup protects against a whole other set of risks: accidental data loss or destruction, theft, fire, explosions, etc.

I find RAID useful.

I don't run systems with RAID (my server supports RAIDing drives but I don't
use it for that) - I have two NASs with RAIDed drives, and a USB RAID drive
enclosure that I can easily move around.

I don't use RAID for performance/size, I use simple mirroring where two drives
(of the same type/size) appears as one, with an automatic backup to the second
drive.

I've always been anal about frequent backups, and RAID give me an automatic
backup of my backups. I swap out the drives every couple years as they fill,
keep the "old" ones for a few more years, and every time I start a new set of
drives I do a full backup to that set, after which I do incremental backups
of added/changed files only - now that I've retired and closed my company, a
drive set goes much longer, but I still like to cycle them every few years.

So far it's worked very well for me .. only a few times (in about 40+ years
have I had to go to backups, and in all that time there's only been 1 (or maybe
2) times a backup drive had died and I needed the RAID - but it was well worth
it to me! (Once set up, there's no more work than doing backups to single
drives)

RAID is useful, and still used, for NAS and SAN applications.

The 'old days' of a RAIDed SCSI chain for bulk local storage is rather outdated, but you totally could with some HBAs and some second hand SAS shelves from eBay.

These days, the only reason to do that is to DIY a more robust NAS with.

If you do that, and use RAID4 with Double Parity, you have a minimum raid group size of 5 disks, but can lose up to 2.

If you are doing a virtual server side-hustle, like hosting game servers, having such a rig, and supplying iSCSI LUNs to the VMs, having them hosted by a raid array like this helps keep service disruption down, and would be worthwhile.

Otherwise?

Probably not worth the effort?

wierd_w wrote on Today, 15:24:

If you do that, and use RAID4 with Double Parity, you have a minimum raid group size of 5 disks, but can lose up to 2.

Isn't that RAID6?

Not quite.

For me, yes, but only RAID 1. I just want backups, speed is irrelevant to me. I mean, I grew up on cassettes and floppy disks. 🤣

Internally, I run a single set of rives in RAID 1 for important data I use regularly. My NAS is 2 drives in RAID 1 for data used less, but access is needed or shared. All other long-term backups are just external drives.